01

Real, fiction, and the
line between

Surveillance Perception Study — the convergence test.
Three real capabilities, three dystopian fictions. Matched on consequence, so the only variable is the source.

95%
of fictions judged real
499
digital twins · US gen-pop
3
consequence-matched pairs
02
Brox.AI

What we tested

Each pair holds the capability, scope and consequence constant — only the source changes. Rated blind, then with the source revealed, plus a believability check.

Pair 1

Predictive policing

Real: Palantir

Fiction: Precrime — Minority Report

Pair 2

Always-on camera

Real: Tesla cabin / Sentry cam

Fiction: The telescreen — 1984

Pair 3

Mass public tracking

Real: Flock Safety

Fiction: Public scanners — Minority Report

Design: both items in a pair use the same consequence-bounded language, matched for length and tone — only the named source changes. Measures: comfort (1–5), purchase / sign-up intent, boycott / protest action — each blind and attributed — plus "is this real or fiction?".

Exact wording shown to twins — blind arm (in the attributed arm the source was appended, e.g. "… — Palantir")

Real (Palantir): "Software that brings together police, public and commercial records to identify which people are most likely to be involved in future crime, and refers them to law enforcement to act on before any offence takes place."
Fiction (Precrime): "A system that brings together surveillance and prediction to identify which people are most likely to commit a crime, and refers them to law enforcement to act on before any offence takes place."
Real (Tesla): "A camera inside your car that records the cabin and the street around it continuously while in use, storing footage the maker can access."
Fiction (telescreen): "A camera inside your home that records the room around it continuously, storing footage the authorities can access at any time."
Real (Flock): "Roadside cameras that automatically identify every passing vehicle and log where and when it was seen into a searchable national database, with no way for a driver to opt out."
Fiction (scanners): "Public scanners that automatically identify every passing person and log where and when they were seen into a searchable record, with no way for anyone to opt out."
03
Brox.AI

Finding 1 — the line is gone

Described plainly, dystopian fiction is indistinguishable from documented reality. The public defaults to "this is real."

Share of twins who judged each fiction capability to be a real, existing technology:

Precrime — Minority Report

99.6% said real

Public scanners — Minority Report

100% said real

The telescreen — 1984

84.4% said real

95%
of all fiction judgments
were "this is real"

The real items scored 100% too. So the public confirms the genuine and waves the fictional through alike — one reflex, running in a single direction: "assume it's already here."

04
Brox.AI

Convergence map — all six items

Believability collapses the 2×2: blind, twins cannot sort real from fiction. Every item — genuine or invented — reads as "real."

CapabilitySource revealedTruthJudged "real"
Predictive policingPalantirReal
100%
Precrime — Minority ReportFiction
99.6%
Always-on cameraTesla (cabin / Sentry)Real
100%
The telescreen — 1984Fiction
84.4%
Mass public trackingFlock SafetyReal
100%
Public scanners — Minority ReportFiction
100%
05
Brox.AI

Finding 2 — consequence sets the line, not the source

Blind comfort (1 = very uncomfortable, 5 = very comfortable). Within each matched pair the real/fiction gap is small — and nothing clears the midpoint.

PairRealFictionGap
Predictive policingPalantir — 1.94Precrime — 1.520.42
Always-on cameraTesla — 1.66Telescreen — 1.010.65
Mass public trackingFlock — 1.52Scanners — 1.110.41

Real ≈ 1.71

Average blind comfort across the three documented capabilities — well below the 3.0 midpoint.

Fiction ≈ 1.21

Average across the three fictions. A half-point separates a real surveillance product from an Orwell device — both are rejected.

These are blind scores — no source is shown, so twins are reacting to the capability alone. Match the consequence and the real/fiction gap nearly closes: people respond to what the capability does to them, not to whether it is documented technology or science fiction. (What a named company adds is the next slide.)

06
Brox.AI

Finding 3 — the brand is a permission structure

Naming a real company raises comfort. There is no trust tax — and no crossover. Naming the fiction source does nothing.

Tesla — always-on camera

+0.88

blind 1.66 → named 2.54

Flock Safety — mass tracking

+0.56

blind 1.52 → named 2.08

Palantir — predictive policing

+0.18

blind 1.94 → named 2.12 · the lone near-flat brand — its name carries its own menace

Mechanism: blind, the phrasing sounds anonymously dystopian. A familiar real name says "this exists and is sold" — and that normalises it. Fiction names have nothing to reassure with, so they stay flat (telescreen 0.00, precrime −0.11, scanners −0.01).

07
Brox.AI

What action people would actually take

Blind, for the three real capabilities. The form action takes hinges on one thing — is there a product to refuse?

19× Tesla draws a boycott nearly nineteen times higher than Flock (64.9% vs 3.4%) — on near-identical discomfort. The gap is the product, not the principle.
Real capability Boycott / refuse to buy Post on social media Join / attend a protest Dislike, but no action
Tesla
a car you buy
64.9%3.8%0.0%31.3%
Flock Safety
roadside infrastructure
3.4%28.3%0.2%65.9%
Palantir
government software
7.6%21.0%0.0%67.5%

When there is something to buy

Tesla is a purchase, so two-thirds (65%) say they would simply not buy it. Concern has a direct, low-effort outlet — withholding the sale.

When there is not

You can't boycott a roadside camera network or government software. So action collapses to a social-media post (Flock 28%, Palantir 21%) — while two-thirds dislike it and do nothing. Almost no one protests in person, for any of the three.

08
Brox.AI

Who actually protests

Protest is the rarest action of all — and the say-do gap lives inside it.

1 Across 499 twins and every capability, exactly one person in each arm would attend an in-person protest. Everyone else who "protests" means clicking join an online group.

The one who shows up

The lone in-person protester in both arms is the same profile: an older man (65+), values-driven, distrustful of institutions — and already in the habit (the reasoning notes he has recently attended "No Kings" protests). He avoids social media, so his action runs offline.

"It feels like Big Brother and it'll hit minorities and poor people first. I'd probably attend an in-person protest, but I'm not gonna be posting on social media."

— male, 65+

The ones who "join a group"

Online-protest selections are almost all women aged 30–64, progressive, who pick the lowest-friction option and explicitly rule out showing up — citing mental health, chronic illness or single-parent time pressure. Naming the brand slightly increases online protest (5 → 7 selections).

"I'd probably just post about it or join some online group, but honestly I'm sick a lot so I wouldn't go do a real protest."

— female, 30–44

09
Brox.AI

Who is more comfortable — and who spots the fiction

Comfort has a clear age and gender gradient. Believability has none.

Comfort rises with youth

Blind comfort with the real capabilities (1 = very uncomfortable, 5 = very comfortable).

Gen Z1.95
Millennials1.80
Gen X1.65
Baby boomers1.66
Silent generation1.33

Men sit consistently above women — 1.83 vs 1.65. No group clears 2.0 on a 5-point scale.

No one can spot the fiction

Share calling the fiction capabilities "real", by generation. The convergence is total across every age group.

Gen Z93.0%
Millennials94.4%
Gen X92.4%
Baby boomers97.5%
Silent generation91.7%

Baby boomers are the most likely to assume fiction is already real. Youth is no defence.

10
Brox.AI

In their words

What the twins say when they explain themselves.

What drives the discomfort

Share of comfort responses raising each theme (n = 2,994).

"Creepy" / uncomfortable87%
Privacy / intrusion70%
Distrust / abuse / misuse65%
Safety / catching crime57%
No choice / no opt-out32%
Government / "Big Brother"27%
40%
of "it's real" judgments said the technology already exists

"I think it's real, like they probably already do that with cameras and data… feels like people could get in trouble for nothing."

— on the Minority Report precrime system (judged real)

"It feels like some Minority Report stuff, and I don't really trust a system to 'predict' who's gonna do a crime and send cops after them."

— recognises the film, still rates the capability real

"I'm not really okay with being tracked in public with no opt out, it feels like too much and easy to misuse."

— on mass public tracking

11
Brox.AI

What it means

Three takeaways for anyone naming, building or regulating surveillance capability.

On attribution

A familiar real-company name acts as permission, not liability — it moves acceptance up. The risk is normalisation, not reputational backlash, except where the name itself carries dread (Palantir).

For the line

"Dystopian" is a wording effect. Once a capability is described in neutral product language, the public can no longer tell it from science fiction — and assumes it is real.

For accountability

Stated concern does not predict action. Infrastructure and B2G surveillance face almost no consumer pressure, regardless of how uncomfortable people are.

The public draws its line by consequence, not by source — and once the consequence is plainly stated, it assumes the line was already crossed.

12
Brox.AI

Method & reading notes

Design

499 Brox digital twins, US general population. Three real/fiction pairs matched on capability, scope and consequence. Each twin rated every item blind, then with the source named — a within-twin reveal, so each twin is its own control.

Measures

Comfort (1–5), purchase / sign-up intent, boycott / protest action, and a real-or-fiction believability check. Blind phrasings are length- and register-matched within each pair.

Caveats

No benefits are stated for any capability, so all comfort scores sit low — read the gaps, not the absolute levels. Within-twin reveal can carry a small re-exposure effect, but the brand halo runs opposite to that bias, so it is robust.

Containment

Every capability is either documented-real or genuine fiction sourced to its work. Nothing is invented or falsely pinned to a real company.

Brox.AI · Surveillance Perception Study · consequence-matched convergence run · 2026-06-30. Figures are digital-twin simulated and intended as a framework read, not fielded human prevalence.